Differentiating Non-O157:H7 STEC Serogroups from Ground Beef Plated on Agar Media by Hyperspectral Imaging

Introduction: The development of an assay to detect and confirm a positive non-O157:H7 isolate is challenging when mixed morphologically results are obtained from the serogroups growing on Rainbow agar.  Rainbow agar is only claimed by the manufacturer to be very specific for E.coli O157:H7 strains which produce black colonies.  It is a bonus that non-O157 serogroups also produce colonies of various colors.  Non-O157 serogroups can produce pink, purple, blue-purple, gray, gray-blue or gray-purple colonies which can be indistinguishable from each other and other background flora.  The challenge for a microbiologist is selecting well isolated non-O157 STEC colonies from enriched samples plated on Rainbow agar for further testing and confirmation. 

Purpose: To investigated the ability of hyperspectral imaging in differentiating non-O157:H7 STEC serogroups (O26, O45, O103, O111, O121, and O145) from ground beef background flora, serial dilutions of enriched ground beef samples spiked with STECs were spread onto agar plates for imaging.

Methods: Ground beef (65+2 g) was enriched in 585 +15 ml of modified tryptic soy broth (m-TSB, 20 mg/ml novobiocin) over night at 42 °C.  Four ten-fold dilutions of the enriched ground beef sample were prepared in sterile saline.  Then for each serogroup, approximate 1,000 CFU (10 ml of a 10⁵ CFU/ml cell suspension) of STEC was spiked into 990 ml of each of the four enriched ground beef sample serial dilutions.  The STEC spiked ground beef sample dilutions were thoroughly mixed then 50 and 100 ml of each dilution was spread onto individual Rainbow agar plates. The Themis Vision Systems’ hyperspectral imaging system was used to acquire images from 400 nm to 900 nm. Regions of interest associated with non-O157 STEC colonies and background flora were created for validation using a previously developed Mahalanobis distance classifier.

Results: PCA score plots revealed potential separability serogroups from each other and the background flora. The prediction with six PCA components showed an overall detection accuracy of 94%. Detection accuracy varied from 88% to 100%.  Sensitivity and specificity of hyperspectral imaging in detecting the target organism and differentiating the target from the background flora was 93% and 100%, repsectivley.

Significance: The potential of combining hyperspectral imaging and chemometrics to differentiate non-O157 STEC serotypes from background flora was demonstrated. Hyperspectral imaging can improve the speed and accuracy of selecting well isolated non-O157 STEC colonies from enriched samples plated on Rainbow agar for further testing and confirmation.